YOUR PERSISTENT COWORKERS
Character studio
Start with a familiar face. Make it their own.
A shared foundation.
A personality of their own.
Prompt a direction, review each stage, and publish
only the character you love.
Your characters
Private until promotedONE SHARED CHARACTER SYSTEM
Base versions
Inspect the originals or branch a new proportion and style preset.
Prompted base versions preserve the shared mesh, skeleton and expression interface. New topology and modular geometry still come from the reproducible Blender pipeline.
READY TO MEET THE WORLD
Published characters
Each release is an immutable snapshot of an approved revision.
ONE AVATAR, ANY WEBSITE
Bring a coworker along.
Promote a reviewed character, then copy its manifest URL. The public manifest contains its appearance, motion and voice configuration. Private prompts and drafts stay in the studio.
Display a character
Connect OpenAI realtime speech
Your website’s server creates an ephemeral OpenAI client secret using its own API key. Pass that short-lived secret to the avatar. Never put your standard OpenAI key in browser code.
await avatar.connectOpenAI({ ephemeralToken });
// End the call and release the microphone:
avatar.disconnect();The SDK exposes setExpression(name, intensity), playAudio(url), attachAudio(mediaStream), metrics() and dispose(). Microphone access requires HTTPS and a user gesture.
Your server: mint a token
Speech uses smooth audio-reactive mouth motion, with conversational expressions selected by OpenAI. It is not phoneme-aligned lip sync. External embeds never spend this studio’s OpenAI key.
OpenAI WebRTC documentation ↗Shape their character.
Each prompt creates a saved branch from the selected revision. Image generation can take a few minutes.
PNG, JPEG or WebP · up to 2 MB. Uploading creates a new branch.
Company background · optional
Only your supplied logo appears. No placeholder or logo is generated.
Fitting notes & genome
LOOK. LISTEN. MAKE IT YOURS.
Meet your coworker.
Uses your microphone and the studio’s OpenAI account. Audio is AI-generated. Mouth motion follows speech energy; phoneme alignment is approximate.
Publishes an immutable avatar reference. Further edits remain private until you approve another release.
EVERY DIRECTION, PRESERVED
Version history
Select any revision to inspect it or create a new branch.